If you are running WCG projects only, BOINC estimating is fairly close.

A mixture of projects can throw the estimating off. Rectilinear Crossing is one that is way out of whack. I loaded a few workunits that estimated over 1.5 hours each and "overcommitted" one computer. WCG workunits are still running based on the overall estimates and the queue is getting smaller than usual. From experience, most of these Rectilinear Crossing workunits will run in times ranging from TWO SECONDS to TWO MINUTES, not 1.5 hours. A giant amount of work will then be needed to fill up the queue again. The longest running RC workunit I have ever seen on this computer is 20 minutes.

UPDATE: It had blasted through the Rectilinear Crossing workunits, and when I got back to it, the queue was empty. Uploads/downloads are taking place now for both projects, and WCG is running.

That's not really necessary. As long as the computer is overcommitted, nothing will happen, but they will be downloaded when the queue is worked down. It is also possible to change the BOINC percentage used for each project but the bad estimates defeat most of the proportions. If WCG is set to no new work, the other one naturally gets 100% eventually.

I am using BOINC version 5.4.11, the current production release. Some of the downloading was done with the previous 5.4.9.

I am not getting millions of seconds of work, because the estimates I have seen are high, not low, as yours must be. Again, a lot of this depends on the project. The sum of the estimated hours is what BOINC is using to compare to the size of the queue.

When workunits have been waiting for a time longer than everything else in the queue, the system will switch to running the oldest workunits first to get them done.

It's still best to finish the WCG workunits in a few days, since the reporting time is set to 7 days. There may be a grace period, but I'm not pushing it.

NOTE: What is your queue length set to? Default time between network connections is 0.1 days and should not be more than a very few days to get WCG work completed.

Abort them now. Then reset the project.

It sounds like something has gone wrong with one of your state files.

It's finally gone in to EDF. I think the BOINC Manager restarted. I'll wait and see how it pans out. If it looks bad I'll start aborting the ones that wont finish in time.

Thanks

Mike

Suspect a conflict between the HDC restricted work availability, the ever increasing request for seconds of work to meet queue settings and not getting enough to fill the queue....then, when it finally runs completely out in the work-buffer for the project, it getting too many upon return of availability.

As written i had 3.2 million seconds of work request (several times) and got only 2 HDC...eventually it said it ran out and gave me the max of FAAH thru the 'alternate work' option as only available for WCG BOINC profiles....I've aborted a few at the end of the queue to not let them expire and force a redistribution too others....the last one's are going to be borderline....we'll see when we get there......unfortunately BOINC will start WU's even if the timestamp indicates expiration...it only warns in the message log.

ciao